The End-to-End CV Framework (EECVF) goal is to offer a flexible and dynamic tool for researching and testing concepts without the need for the user to handle the interconnections through the system. To better overcome the continuous development of the EECVF it is constructed as a modular and scalable concept.
The term “End-to-End” describes in our case the ability of the framework to execute several stages of a CV process as a “one click” solution. The framework we propose has the capability to create-train-evaluate a ML model, run a CV application using the model, evaluate the results and plot the results without any intervention of the user. All of the steps being done in parallel with documenting debug information desired by the user.
The framework’s main goal is to unify different stages of the research from the CV vast domain. If we look in Figure 1 we can see the blocks that form the framework. Reduction of redundant operations and calculations done by the system is one of the benefits of the “one click” solution. Another benefit generated by the design of the framework is the lack of duplication of data or interfaces through the system. A user can only use one of the blocks, for example only the Application block to run simple CV pipeline or just the Benchmark block to evaluate results.
Each block has an interface module that will present the jobs and services that exposed in each block. This mechanism permits the isolation between the user experience and the inner works of each block. This is an important feature of EECVF because it creates an environment in which the users would only focus on the research topic at hand and not on the tools they have to use.